Methicillin-resistant S. aureus strains included in this study were isolated from routinely sampled sputum specimens in patients attending the CF center at Montpellier University Hospital, France, and analyzed as part of their standard follow-up. Strains were recovered in 35 patients representing all patients with MRSA isolation in our center over a 9-year period. Strains had been stored at −80°C. These strains were studied for their ability to form biofilm in the presence or absence of antibiotics according to the steps presented in the flow diagram in Figure 1. Briefly, we first identified strains that were strong biofilm producers among a collection of clinically documented and genetically characterized CF MRSA using the BRT® and the ASM. Strong biofilm producers were then included in the evaluation of antibiotic efficacy against biofilm formation using the Antibiofilmogram® method. Selected isolates were then used to conduct: (i) a longitudinal analysis of the evolution of antibiotic efficacy on biofilm formation over time in chronically colonized patients (selection of clonal strains successively isolated in four patients) and (ii) a study of the dynamics of biofilm formation under flow in the absence and presence of antibiotics (selection of two paired strains successively isolated in a chronically colonized patient).

All strains were cultured on Trypticase Soy Agar with 5% Sheep Blood (bioMérieux, Marcy l’Étoile, France), and one colony was subcultured on the same agar medium for further analysis. DNA extraction was performed as previously described (Predari et al., 1991). The amplification of the seven housekeeping genes included in the MLST scheme for S. aureus was performed as previously described (Enright et al., 2000; Crisóstomo et al., 2001). Sequencing was done on a 3500xL Genetic Analyser (Thermo Fisher Scientific Waltham, Massachusetts, United States), and sequence analysis was performed using SeqScape Software v3.0™ from the same company. The sequence type (ST) was determined for each strain via the PubMLST database (Jolley et al., 2018)¹ or after whole-genome sequencing (wgMLST).

Whole-genome sequencing analysis was performed for the 11 CF MRSA isolates (five “early” isolated strains and six “late” isolated strains in four chronically colonized patients) selected for the longitudinal analysis of the evolution of antibiotic efficacy on biofilm formation over time to prove that the MRSA persistence in these patients was truly a chronic colonization by a clonal strain. For six of these strains, WGSs were available from previous work (Boudet et al., 2021); for the five remaining strains, WGSs were obtained by an Illumina MiSeq platform (Illumina Inc., San Diego, CA, United States) and analyzed as previously described (Boudet et al., 2021). Regarding analysis of single nucleotide polymorphisms (SNPs), SNPs were called using Snippy (Seemann, 2015) and SNPs numbers were interpreted according to the criteria of Ankrum and Hall (2017) which define strains with ≤71 SNPs as the “same” strains, strains with 72–123 SNPs as “very closely related” strains, strains with 124–156 SNPs as “closely related” strains, and strains with ≥157 SNPs as “distantly related” strains.

DDBJ/ENA/GenBank accession numbers for the 11 WGS are JAGPWS000000000, JAGPWY000000000 and JAGPWZ000000000 (strains 5.1, 5.2, and 5.3 in Patient 5); JAGPWI000000000, JAGPWQ000000000 and JAGPWR000000000 (strains 6.1, 6.3, and 6.4 in Patient 6); JAHXBM000000000, JAHXBL000000000, and JAHXBK000000000 (strains 17.2, 17.3, and 17.6 in Patient 17); JAHXBJ000000000 and JAHXBI000000000 (strains 18.1 and 18.2 in Patient 18).

After initial culture of frozen strains on Trypticase Soy Agar with 5% Sheep Blood (bioMérieux, Marcy l’Étoile, France), we used different media according to the test performed. Mueller-Hinton broth was used as the broth microdilution reference method to determine minimum inhibitory concentrations (MIC) of antibiotics. The ASM, a medium mimicking the sputum found in the respiratory tract of CF patients (Sriramulu et al., 2005; Dinesh, 2010), was used to evaluate biofilm formation and antibiotic efficacy against biofilm formation in conditions more relevant to the CF lung using the different approaches of this study and developed below: the Biofilm ring Test® (BRT®), the Antibiofilmogram®, and the continuous flow system BioFlux™ 200. However, the brain heart infusion (BHI) medium is recommended by the BRT® manufacturer and was therefore used with ASM for comparative purposes. As the opacity of the ASM did not provide a compatible image with the BRT® reading software, 0.22μm-filtered ASM was compared to BHI medium on selected isolates with the BRT®. BHI medium was also compared with ASM in the assays conducted on the BioFlux™ system. According to the medium used for the assay, strains were pre-cultured for 24h in the same medium, BHI broth, or ASM. Due to the use of these three culture conditions, growth properties of selected strains in BHI broth, ASM, and 0.22μm-filtered ASM were compared. Overnight cultures at 37°C were inoculated into the same fresh medium [identical dilution factor for all three conditions to obtain a suspension of optical density (OD) at 600nm (OD₆₀₀) of 0.1], and incubation was carried out at 37°C with agitation for 24h. Cell growth was monitored over 24h through colony-forming unit (CFU) counts with eight data points: 0, 0.5, 1, 2, 3, 4, 6, and 24h. For CFU counts, cultures were diluted serially in the medium and plated on Trypticase Soy Agar with 5% Sheep Blood plates which were incubated for 24h at 37°C. All assays were performed in duplicate.

The BRT® based on the measurement of the mobility of superparamagnetic microbeads subjected to a magnetic field was used to study the early stages of biofilm formation according to the manufacturer’s recommendations (Chavant et al., 2007). Briefly, standardized bacterial cultures were incubated at 37°C in 96-well microtiter plates in the presence of magnetic beads. At set time-points, the plates were placed on a magnetic block and put in the reader. The images of each well before and after magnetic attraction were analyzed with the BioFilm Control software (BFC Elements® 3), which gives a biofilm formation index (BFI). The adhesion ability of each strain was expressed as this BFI that is inversely proportional to the attached cell number. A high BFI value indicates high bead mobility under magnetic action that corresponds to the absence of biofilm formation (a spot is visible due to bead accumulation above the mini-magnet), while lower values reflect partial to complete immobilization of beads due to the bacteria embedded in a biofilm (no visible spot for complete bead immobilization). Two incubation times were defined: 1.5 and 4h. Three replicate wells were performed per strain and per incubation time in three (BHI/filtered ASM comparison) or two independent experiments (tests in filtered ASM only). A negative control was systematically included in each experiment corresponding to the medium and beads without bacterial suspension. BFI values were calculated for each well, ranging from 0 (total bead immobilization, i.e., strongly adherent cells/strong biofilm formation) to 20 (no bead aggregation, i.e., non-adherent cells/no biofilm formation in the experiment conditions).

Minimum inhibitory concentrations of ceftaroline, ceftobiprole, linezolid, rifampicin, and trimethoprim were determined for all CF MRSA strains using the microdilution method in Mueller-Hinton broth. According to the interpretative criteria of the European Committee on Antimicrobial Susceptibility Testing (EUCAST), strains were classified as susceptible (including “susceptible” and “susceptible, increased exposure” categories) or resistant (European Committee on Antimicrobial Susceptibility Testing, 2021). Serial 2-fold dilutions of antibiotics were as follows: 0.125–16mg/L for ceftaroline, 0.25–32 for ceftobiprole, 0.5–64mg/L for linezolid, 0.015–2mg/L for rifampicin, and 0.25–32 for trimethoprim. The reference strain S. aureus ATCC 29213 was used to monitor test performance.

The minimum biofilm inhibitory concentrations (bMIC) were determined using the Antibiofilmogram® test (BioFilm Control; Tasse et al., 2016). A schematic representation of the Antibiofilmogram® principle is available in Olivares et al. (2020). The test was performed using 0.22μm-filtered ASM for initial bacterial growth and suspension preparation. The microplates containing bacteria, magnetic beads, and antibiotics (20μl of antibiotic solutions) were incubated at 37°C for 4h before visual reading. The bMIC was determined for each antibiotic as the lowest concentration at which a spot corresponding to free beads attracted to the center of each well during magnetization, similar to negative control (0.22μm-filtered ASM and magnetic beads), was visible. A well without antibiotics filled with the bacterial suspension in 0.22μm-filtered ASM and magnetic beads was used as the positive control (absence of spot due to bead immobilization in biofilm). All assays were performed in duplicate.

The kinetics of biofilm formation and quantification in the absence and presence of antibiotics (ceftaroline, ceftobiprole, and linezolid) were assessed with the BioFlux™ 200 microfluid system as previously described (Naudin et al., 2019) with two different media, BHI medium and ASM. In this system, biofilm formation is studied under dynamic and controlled flow conditions and is followed by light microscopy in microfluidic wells (Benoit et al., 2010). The system consists of a 48-well plate with a microchannel connection between 24 pairs of two types of wells, an input well and an output well. Bacteria were grown overnight in each media. Bacterial suspensions with an OD₆₀₀ of 0.1 +/− 0.05 after two 200th serial dilutions from the overnight culture were prepared in both media. To quantify biofilm formation, we prepared BioFlux system by adding 500μl of medium into the input well with a pressure of 1 dyne/cm² for 10min. At the end of the 10min, the medium remaining in the input well was removed. Bacterial suspensions were then added in the input well for 36h of incubation at 37°C in BHI medium and ASM with a pressure of 0.2 dyne/cm². The bacterial suspension was either added alone or in the antibiotic solution, i.e., last dilution performed in the antibiotic solution to give final concentrations of bacteria of 10³CFU/ml and antibiotic corresponding to the bMIC determined by the Antibiofilmogram® either in BHI broth or in filtered ASM. Biofilms were grown in monoculture, and the quantification of biofilm formation was done at 5, 12, 24, and 36h post-inoculation in the absence of antibiotics and at 5, 12, and 36h post-inoculation in the presence of antibiotics. Experiments were performed in duplicate. Biofilms were visualized with a Leica DM IRB inverted fluorescence microscope coupled with a CoolSNAP FX black and white camera (Roper Scientific, Trenton NJ, United States). ImageJ® software was utilized to calculate biofilm percentage. The 16-bit grayscale images were adjusted with the threshold function to fit the bacterial structure and were analyzed using the “Analyze Particles” function to calculate biofilm percentage in the microfluidic channel.

Results are presented as the mean±standard deviation. Statistical analyses were made using a t test in version 7 of GraphPad Prism (comparison of biofilm formed according to the condition, either medium or incubation time, in the BioFlux™ analyses) or using a Mann-Whitney U test in version 3.5.2 of R software (comparison of BFI values in BHI medium and filtered ASM). A value of p≤0.05 was considered to reflect significance.

The 35 CF patients (54.3% of female) had a mean age of 22.1years at inclusion in the study (range: 2–52years). Sixteen of them fulfilled the criteria for chronic colonization (>50% of respiratory samples collected during the last 12months positive for MRSA with at least four samples collected during that period; Zolin et al., 2020). The 63 strains had been isolated between 2007 and 2016 from 35 sputum samples in these patients (2–4 colonial morphotypes visually distinct observed during the routine bacteriological analysis of sputum samples recovered from 18 samples in 18 patients were included; Supplementary Table). MLST genotyping showed that the 63 strains belonged to 10 STs (Figure 2A). MRSA of ST5 (50.8% of strains and 48.6% of patients) and ST8 (30.2% of strains and 31.4% of patients) were the most frequent representing 81% of the strains (51 strains/63) and being identified in 80% of the patients (28 patients/35). Two patients had co-colonizations with strains from different STs within the same clonal complex (CC) identified in one sample [ST5 and ST4782 (CC5) in Patient 14, ST8 and ST5829 (CC8) in Patient 33]. In these cases, both genotypes identified differed by one (strains in Patient 14) or five mutations (strains in Patient 33) in one of the seven genes of the MLST scheme (aroE gene and gmk gene, respectively).

Brain heart infusion medium is recommended by the BRT® manufacturer. However, the ASM containing mucin, amino acids, and free DNA was formulated to mimic the sputum of CF patients and its use is thought to be more representative of the lung conditions during CF. We therefore aimed to evaluate the use of ASM in the Biofilm Ring Test®, either to characterize the adhesion pattern and ability to form biofilm of clinical CF MRSA (BFI determination) or to perform an Antibiofilmogram® (bMIC determination). ASM is a turbid medium which yields invalid results in the BRT® but using ASM filtered through a 0.22μM filter system made it possible for the system to calculate the BFI. To go further with the use of filtered ASM, we compared the growth ability in BHI, ASM, and filtered ASM of three selected strains that were further categorized as strong biofilm producers: strains 5.3 and 6.4 that were also used for the comparison of the BRT® use in BHI medium and ASM, and strain 17.6 that was later studied using the BioFlux™ 200 system. Similar growth rates were observed in the three media for these three strains (Supplementary Figure 1). We then compared both media for BRT® use by comparing BFI values measured after 4h of incubation in BHI medium and filtered ASM for strains 5.3 and 6.4 (Supplementary Figure 2). Differences in adhesion patterns were observed depending on the medium used, with lower BFI values observed in filtered ASM compared with BHI medium and a significantly more pronounced adherence behavior in ASM compared with BHI medium for strain 5.3 (value of p<0.001 by Mann-Whitney U test). On the whole, these observations led us to consider the use of filtered ASM for further experiments using the BRT®.

Biofilm formation index values evaluating the CF MRSA adhesion ability were distributed in three clearly separated groups after an incubation time of 4h: BFI values≤4, BFI values comprised between 8.37 and 13.52, and BFI values ranging from 19.87 to 20.31 (Supplementary Table; Supplementary Figure 3). We found that 27% of strains (17/63) isolated in samples from 37.1% of the patients (13/35) formed a strong biofilm after 4h of incubation in filtered ASM (BFI<4), including two strains of ST5 (strains 17.6 and 30.1) which were strongly adherent after 1.5h of incubation (Figure 2B). Nearly all the strains (94.1%, 16/17) that were early biofilm producers belonged to ST5 and ST8. It is worth noting that intra-sample heterogeneity in the ability to form biofilm was observed, strong biofilm-producing strains being co-isolated with strains displaying distinct ability to form biofilm in samples from five patients (38.5%, 5/13). A striking example was observed for Patient 25 for whom strong biofilm producers and non-adherent strains coexisted in a sample (Supplementary Table). Half of chronically colonized patients (8/16) were colonized by such strains with a strong capacity for biofilm production (Supplementary Table). Non-adherent strains represented 44.4% of the strains and were isolated in 18 patients (BFI: 19.87–20.31, as observed for the negative control that contained filtered ASM and beads only). Finally, 28.6% of the strains, isolated from 13 patients, with intermediate BFI values (8.37–13.52), were likely slower biofilm formers, and characterization of their adhesion kinetics would have required a longer incubation time (Supplementary Table; Supplementary Figure 3).

No obvious relation was observed between the ability to form early biofilm of the MRSA strains and the co-colonizers, particularly Pseudomonas aeruginosa. Out of the 63 strains/35 patients of the study, 40 MRSA strains were isolated in 21 patients with chronic colonization by P. aeruginosa. These strains were distributed in the three groups of strains defined according to their ability to form early biofilm (13 strains showed BFI≤4, 17 strains were non-adherent and 10 had intermediate BFI values).

Determination of the MICs of the five antibiotics under consideration in this study showed that 11 strains (64.7%) were resistant to rifampicin, and three strains (17.6%) were resistant to linezolid, while the 17 strains were susceptible to trimethoprim, ceftobiprole, and ceftaroline (Table 1). bMICs were then determined using the Antibiofilmogram® approach in filtered ASM. Highly distinct patterns of bMICs were observed according to the antimicrobial agent (Table 1). No strains displayed bMIC for trimethoprim below the EUCAST threshold for resistance at 4mg/L (bMIC range: 8–>32mg/L) showing that this antibiotic was totally inefficient on biofilm formation initiation. The lowest bMICs were observed for ceftaroline and ceftobiprole with efficacy in preventing biofilm formation observed for 76.5 and 70.6% of strains, respectively (bMICs below the respective resistance breakpoint of 1 and 2mg/L) compared with linezolid (64.7% of strains with bMICs≤4mg/L) and rifampicin (17.6% of strains with bMICs≤0.5mg/L). Individual results for each of the 17 strains under evaluation are presented in Table 2. The capacity of rifampicin and linezolid to prevent biofilm installation was also evaluated for the sole strains that were susceptible in planktonic cultures; linezolid was able to inhibit biofilm formation of 78.6% of the susceptible strains (three out of 14) and rifampicin prevented biofilm installation of 50% of the susceptible strains (three out of six; Table 2).

For three strains (strains 1.1, 5.3, and 20.1), bMICs of all the five antibiotics were above the corresponding resistance breakpoint and these strains were excluded from further analysis conducted to compare the efficacy of the various antimicrobial agents on biofilm formation by CF MRSA. Analysis was, thus, restricted to susceptible strains (planktonic cultures) that displayed at least one bMIC below the threshold for resistance (n=14; Table 2). Under these conditions, bMICs for ceftaroline were below the resistance threshold for 13 strains, bMICs for ceftobiprole were below the resistance threshold for 12 strains, and bMICs for linezolid were below the resistance threshold for 11 strains. A limited effect was observed for rifampicin with three strains out of 14, isolated from three patients, showing low bMIC values<0.015mg/L. However for two patients, these strains were associated either with a resistant strain (Patient 28) or with a strain for which the formation of biofilm was not inhibited by rifampicin (Patient 25) thereby making the drug of low interest considering its global effect on biofilm formation inhibition. More generally, intra-sample heterogeneity with coexistence of adaptive variants displaying distinct bMICs was observed for three out of the four patients for whom the analysis of multiple strains from a sample was performed (Patients 5, 25, and 28; Table 2).

For chronically colonized patients, we searched for strains that had been isolated before the strains studied above, which met the following criteria: (i) strongly adherent strains/strains forming early biofilm as determined by the BRT® in filtered ASM (BFI≤4), and (ii) WGS-based analysis confirming the chronic colonization based on the identification of less than 123 SNPs between the sequences of the “early” and the “late” isolated strains (signing either a “same” strain or “very closely related” strains according to the criteria of Ankrum and Hall, 2017). A total of five earlier-isolated strains from four patients matching these criteria were selected and compared with six strains isolated during patient follow-up. The characteristics of patients and selected strains are presented in Figure 3A. MICs were determined, and an Antibiofilmogram® was performed on these strains (as described above), and bMICs are presented in Figure 3B. MIC and bMIC values for the five antibiotics were all below the EUCAST threshold for susceptibility for the five strains isolated earlier. For later-isolated strains that remained susceptible to antimicrobial agents in planktonic culture, we showed that ceftaroline and ceftobiprole (Patients 5 and 6) or linezolid (Patients 17 and 18) retained their ability to prevent biofilm formation (bMICs below the corresponding EUCAST thresholds for resistance) over time. We related these observations to the antimicrobial courses received by the patients (Figure 3). The more important antibiotic selective pressure was noted for linezolid and Patients 5 and 6. Three out of the four strains isolated in these patients were resistant to linezolid in planktonic cultures (resistance, investigated in a previous study; Boudet et al., 2021) was related to a G2576T substitution present in a variable number of 23S rRNA gene copies, while susceptible strain 6.3 displayed a high bMIC of 16mg/L. For strains isolated in Patients 17 and 18 (who had received fewer linezolid courses during their follow-up), increased bMICs were observed over time, but bMICs were still below the resistance threshold (Figure 3). Although a low number of rifampicin and trimethoprim courses (possibly associated) were noted in the four patients, both antibiotics were no more active on later-collected strains, showing MICs or bMICs above the corresponding resistance thresholds. None of the patients had received ceftaroline or ceftobiprole.

To complete the characterization of biofilm formation of CF MRSA in the presence or absence of antibiotics, a study was performed in dynamic conditions using the BioFlux™ 200 microfluidic system. We selected strain 17.6, one of the two strains that were strongly adherent after 1.5h of incubation in the BRT® and the paired 17.2 strain isolated 3years before in Patient 17. For these paired strains, we first compared their dynamics of biofilm formation over time according to the medium used (BHI or ASM). The formation of biofilm increased regularly over time for each condition of the assay (“early” strain in BHI medium, “early” strain in ASM, “late” strain in BHI medium, “late” strain in ASM) with a systematically higher percentage of biofilm formed in ASM compared with BHI medium and for “late” strain compared with “early” strain (Figure 4). Although both strains were shown to have high adhesion ability using the BRT®, these results revealed that strains 17.2 and 17.6 each had distinct dynamics of biofilm formation, with a slower biofilm formation for the earlier-isolated strain 17.2 (Figure 4).

Further testing with antibiotics was limited to three times of measure (5, 12, and 36h) and to the three antimicrobial agents with the lowest bMICs, that is, ceftaroline, ceftobiprole, and linezolid. With the BioFlux™ device, the percentage of biofilm formed in the presence of one of the three drugs tested was drastically lower than that observed for the control without antibiotics showing that all three antibiotics had the ability to limit biofilm formation and that this effect was long-lasting as it was observed up to 36h of exposure to the bMIC of the corresponding antimicrobial agent in both media (Figure 5). Differences in biofilm formation in ASM were compared for statistical significance by a t test. A significantly higher inhibition of biofilm formed by the earlier isolated strain 17.2 was observed from 12h of incubation and up to 36h of incubation for ceftaroline and ceftobiprole compared with linezolid (values of p<0.05 at 12h; values of p<0.01 at 36h). Differences observed between ceftaroline and ceftobiprole were also statistically significant (value of p<0.05 at 36h), ceftaroline being more active in biofilm formation inhibition than ceftobiprole after 12h, whereas ceftobiprole was more active in biofilm formation inhibition than ceftaroline after 36h. Inhibition of the biofilm formation by the “late” strain 17.6 in ASM was more pronounced with ceftobiprole than with linezolid (values of p<0.01 at 12 and 36h) or ceftaroline (values of p<0.01 at 36h). Ceftaroline was also more active than linezolid after 36h (value of p<0.01).